Every gene in the H2B ubiquitination axis, ranked against the whole genome for KMT2A selectivity — and the writers are not selective

🟠 **ARM-COMPOSITION CAVEAT ADDED 2026-09-24 (cycle 140).** This entry's computation rests on `analyses/depmap-kmt2a-selective-scan-full`, and that scan **fixes both comparison arms from the first gene that returns screen data** (`run.py:170`, `if rear is None:`; line 174 prints *"Groups fixed from first gene with data"*). Because DepMap screens each gene in a gene-specific set of lines, **5 of 16 eligible KMT2A-rearranged lines are excluded from every gene comparison, including genes where they were screened** - 16 resolve under the scan's own query, 11 were used. **This does not show any conclusion here is wrong.** Dropping 5 of 16 costs power, and it biases a ranking only if the dropped lines differ systematically from the kept ones, **which nobody has tested.** The named one-line fix is to derive the arms from each gene's own screen list. Until the scan is re-run, selectivity ranks taken from it carry an unquantified arm-composition caveat. Found by `analyses/depmap-scan-line-identities` on 2026-09-21 and unreported until now; see `ana-the-depmap-selectivity-scan-fixes-its-arms-from-the-first-gene-and-drops-five-of-sixteen-lines-from-every-comparison`. --- The premise of this beat is that H2B ubiquitination is upstream of the most-cited mechanistic step in KMT2A-rearranged leukaemia and absent from the graph. The premise is correct. The natural corollary — that the machinery is therefore a target — is what this analysis tests, using the project's existing full genome-wide scan (`analyses/depmap-kmt2a-selective-scan-full`, 17,728 protein-coding genes, 11 Cellosaurus-annotated KMT2A-rearranged versus 43 other leukaemia lines) with no new computation, only lookups of `all_results`. **The writers fail.** RNF20 ranks 16,229 of 17,728 — its difference is *positive*, meaning marginally less required in the rearranged lines — while being broadly essential (-0.734). RNF40 (rank 9,953), WAC (7,779), UBE2A (6,504), UBE2B (10,016), UBE2E1 (17,038) and MSL2 (7,112) are all flat. **The erasers do better.** USP22 ranks 226 and ATXN7L3 ranks 290, both in the top 1.6% of the genome, but ENY2 — ATXN7L3's own module partner — ranks 11,080, which is an internal inconsistency that argues for caution. For scale, MEN1 ranks 1 and DOT1L ranks 184 in the same scan.